source: mainline/uspace/srv/vfs/vfs_ops.c@ 5bcd5b7

lfn serial ticket/834-toolchain-update topic/msim-upgrade topic/simplify-dev-export
Last change on this file since 5bcd5b7 was 5bcd5b7, checked in by Jiri Zarevucky <zarevucky.jiri@…>, 12 years ago

Add support for server-side mounts.
  • Property mode set to 100644
File size: 33.8 KB
Line 
1/*
2 * Copyright (c) 2008 Jakub Jermar
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 *
9 * - Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * - Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 * - The name of the author may not be used to endorse or promote products
15 * derived from this software without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27 */
28
29/** @addtogroup fs
30 * @{
31 */
32
33/**
34 * @file vfs_ops.c
35 * @brief Operations that VFS offers to its clients.
36 */
37
38#include "vfs.h"
39#include <macros.h>
40#include <stdint.h>
41#include <async.h>
42#include <errno.h>
43#include <stdio.h>
44#include <stdlib.h>
45#include <str.h>
46#include <stdbool.h>
47#include <fibril_synch.h>
48#include <adt/list.h>
49#include <unistd.h>
50#include <ctype.h>
51#include <fcntl.h>
52#include <assert.h>
53#include <vfs/canonify.h>
54#include <vfs/vfs_mtab.h>
55
56FIBRIL_MUTEX_INITIALIZE(mtab_list_lock);
57LIST_INITIALIZE(mtab_list);
58static size_t mtab_size = 0;
59
60/* Forward declarations of static functions. */
61static int vfs_truncate_internal(fs_handle_t, service_id_t, fs_index_t,
62 aoff64_t);
63
64/**
65 * This rwlock prevents the race between a triplet-to-VFS-node resolution and a
66 * concurrent VFS operation which modifies the file system namespace.
67 */
68FIBRIL_RWLOCK_INITIALIZE(namespace_rwlock);
69
70vfs_node_t *root = NULL;
71
72#if 0
73static int vfs_attach_internal(vfs_node_t *mpoint, vfs_node_t *mountee)
74{
75 assert(mpoint != NULL);
76 assert(mountee != NULL);
77
78 if (mpoint->mount != NULL) {
79 return EBUSY;
80 }
81
82 if (mpoint->type != VFS_NODE_DIRECTORY) {
83 return ENOTDIR;
84 }
85
86 if (vfs_node_has_children(mpoint)) {
87 return ENOTEMPTY;
88 }
89
90 mpoint->mount = mountee;
91 vfs_node_addref(mountee);
92 /* Add reference to make sure the node is not freed. Removed in detach_internal(). */
93 vfs_node_addref(mpoint);
94 return EOK;
95}
96
97static int vfs_detach_internal(vfs_node_t *mpoint)
98{
99 assert(mpoint != NULL);
100
101 if (mpoint->mount == NULL) {
102 return ENOENT;
103 }
104 vfs_node_put(mpoint->mount);
105 mpoint->mount = NULL;
106 vfs_node_put(mpoint);
107}
108#endif
109
110static int vfs_mount_internal(ipc_callid_t rid, service_id_t service_id,
111 fs_handle_t fs_handle, char *mp, char *opts)
112{
113 vfs_lookup_res_t mp_res;
114 vfs_lookup_res_t mr_res;
115 vfs_node_t *mp_node = NULL;
116 vfs_node_t *mr_node;
117 fs_index_t rindex;
118 aoff64_t rsize;
119 async_exch_t *exch;
120 sysarg_t rc;
121 aid_t msg;
122 ipc_call_t answer;
123
124 /* Resolve the path to the mountpoint. */
125 fibril_rwlock_write_lock(&namespace_rwlock);
126 if (root == NULL) {
127 /* We still don't have the root file system mounted. */
128 if (str_cmp(mp, "/") != 0) {
129 /*
130 * We can't resolve this without the root filesystem
131 * being mounted first.
132 */
133 fibril_rwlock_write_unlock(&namespace_rwlock);
134 async_answer_0(rid, ENOENT);
135 return ENOENT;
136 }
137
138 /*
139 * For this simple, but important case,
140 * we are almost done.
141 */
142
143 /* Tell the mountee that it is being mounted. */
144 exch = vfs_exchange_grab(fs_handle);
145 msg = async_send_1(exch, VFS_OUT_MOUNTED,
146 (sysarg_t) service_id, &answer);
147 /* Send the mount options */
148 rc = async_data_write_start(exch, (void *)opts,
149 str_size(opts));
150 vfs_exchange_release(exch);
151
152 if (rc != EOK) {
153 async_forget(msg);
154 fibril_rwlock_write_unlock(&namespace_rwlock);
155 async_answer_0(rid, rc);
156 return rc;
157 }
158 async_wait_for(msg, &rc);
159
160 if (rc != EOK) {
161 fibril_rwlock_write_unlock(&namespace_rwlock);
162 async_answer_0(rid, rc);
163 return rc;
164 }
165
166 rindex = (fs_index_t) IPC_GET_ARG1(answer);
167 rsize = (aoff64_t) MERGE_LOUP32(IPC_GET_ARG2(answer),
168 IPC_GET_ARG3(answer));
169
170 mr_res.triplet.fs_handle = fs_handle;
171 mr_res.triplet.service_id = service_id;
172 mr_res.triplet.index = rindex;
173 mr_res.size = rsize;
174 mr_res.type = VFS_NODE_DIRECTORY;
175
176 /* Add reference to the mounted root. */
177 root = vfs_node_get(&mr_res);
178 assert(root);
179
180 fibril_rwlock_write_unlock(&namespace_rwlock);
181 async_answer_0(rid, rc);
182 return rc;
183 }
184
185 /* We already have the root FS. */
186 if (str_cmp(mp, "/") == 0) {
187 /* Trying to mount root FS over root FS */
188 fibril_rwlock_write_unlock(&namespace_rwlock);
189 async_answer_0(rid, EBUSY);
190 return EBUSY;
191 }
192
193 rc = vfs_lookup_internal(root, mp, L_DIRECTORY, &mp_res);
194 if (rc != EOK) {
195 /* The lookup failed. */
196 fibril_rwlock_write_unlock(&namespace_rwlock);
197 async_answer_0(rid, rc);
198 return rc;
199 }
200
201 mp_node = vfs_node_get(&mp_res);
202 if (!mp_node) {
203 fibril_rwlock_write_unlock(&namespace_rwlock);
204 async_answer_0(rid, ENOMEM);
205 return ENOMEM;
206 }
207
208 /*
209 * Now we hold a reference to mp_node.
210 * It will be dropped upon the corresponding VFS_IN_UNMOUNT.
211 * This prevents the mount point from being deleted.
212 */
213
214 /*
215 * At this point, we have all necessary pieces: file system handle
216 * and service ID, and we know the mount point VFS node.
217 */
218
219 async_exch_t *mountee_exch = vfs_exchange_grab(fs_handle);
220 assert(mountee_exch);
221
222 exch = vfs_exchange_grab(mp_res.triplet.fs_handle);
223 msg = async_send_4(exch, VFS_OUT_MOUNT,
224 (sysarg_t) mp_res.triplet.service_id,
225 (sysarg_t) mp_res.triplet.index,
226 (sysarg_t) fs_handle,
227 (sysarg_t) service_id, &answer);
228
229 /* Send connection */
230 rc = async_exchange_clone(exch, mountee_exch);
231 vfs_exchange_release(mountee_exch);
232
233 if (rc != EOK) {
234 vfs_exchange_release(exch);
235 async_forget(msg);
236
237 /* Mount failed, drop reference to mp_node. */
238 if (mp_node)
239 vfs_node_put(mp_node);
240
241 async_answer_0(rid, rc);
242 fibril_rwlock_write_unlock(&namespace_rwlock);
243 return rc;
244 }
245
246 /* send the mount options */
247 rc = async_data_write_start(exch, (void *) opts, str_size(opts));
248 if (rc != EOK) {
249 vfs_exchange_release(exch);
250 async_forget(msg);
251
252 /* Mount failed, drop reference to mp_node. */
253 if (mp_node)
254 vfs_node_put(mp_node);
255
256 fibril_rwlock_write_unlock(&namespace_rwlock);
257 async_answer_0(rid, rc);
258 return rc;
259 }
260
261 /*
262 * Wait for the answer before releasing the exchange to avoid deadlock
263 * in case the answer depends on further calls to the same file system.
264 * Think of a case when mounting a FS on a file_bd backed by a file on
265 * the same FS.
266 */
267 async_wait_for(msg, &rc);
268 vfs_exchange_release(exch);
269
270 if (rc == EOK) {
271 rindex = (fs_index_t) IPC_GET_ARG1(answer);
272 rsize = (aoff64_t) MERGE_LOUP32(IPC_GET_ARG2(answer),
273 IPC_GET_ARG3(answer));
274
275 mr_res.triplet.fs_handle = fs_handle;
276 mr_res.triplet.service_id = service_id;
277 mr_res.triplet.index = rindex;
278 mr_res.size = rsize;
279 mr_res.type = VFS_NODE_DIRECTORY;
280
281 /* Add reference to the mounted root. */
282 mr_node = vfs_node_get(&mr_res);
283 assert(mr_node);
284 } else {
285 /* Mount failed, drop reference to mp_node. */
286 if (mp_node)
287 vfs_node_put(mp_node);
288 }
289
290 async_answer_0(rid, rc);
291 fibril_rwlock_write_unlock(&namespace_rwlock);
292 return rc;
293}
294
295void vfs_mount(ipc_callid_t rid, ipc_call_t *request)
296{
297 service_id_t service_id;
298
299 /*
300 * We expect the library to do the device-name to device-handle
301 * translation for us, thus the device handle will arrive as ARG1
302 * in the request.
303 */
304 service_id = (service_id_t) IPC_GET_ARG1(*request);
305
306 /*
307 * Mount flags are passed as ARG2.
308 */
309 unsigned int flags = (unsigned int) IPC_GET_ARG2(*request);
310
311 /*
312 * Instance number is passed as ARG3.
313 */
314 unsigned int instance = IPC_GET_ARG3(*request);
315
316 /* We want the client to send us the mount point. */
317 char *mp;
318 int rc = async_data_write_accept((void **) &mp, true, 0, MAX_PATH_LEN,
319 0, NULL);
320 if (rc != EOK) {
321 async_answer_0(rid, rc);
322 return;
323 }
324
325 /* Now we expect to receive the mount options. */
326 char *opts;
327 rc = async_data_write_accept((void **) &opts, true, 0, MAX_MNTOPTS_LEN,
328 0, NULL);
329 if (rc != EOK) {
330 free(mp);
331 async_answer_0(rid, rc);
332 return;
333 }
334
335 /*
336 * Now, we expect the client to send us data with the name of the file
337 * system.
338 */
339 char *fs_name;
340 rc = async_data_write_accept((void **) &fs_name, true, 0,
341 FS_NAME_MAXLEN, 0, NULL);
342 if (rc != EOK) {
343 free(mp);
344 free(opts);
345 async_answer_0(rid, rc);
346 return;
347 }
348
349 /*
350 * Wait for VFS_IN_PING so that we can return an error if we don't know
351 * fs_name.
352 */
353 ipc_call_t data;
354 ipc_callid_t callid = async_get_call(&data);
355 if (IPC_GET_IMETHOD(data) != VFS_IN_PING) {
356 async_answer_0(callid, ENOTSUP);
357 async_answer_0(rid, ENOTSUP);
358 free(mp);
359 free(opts);
360 free(fs_name);
361 return;
362 }
363
364 /*
365 * Check if we know a file system with the same name as is in fs_name.
366 * This will also give us its file system handle.
367 */
368 fibril_mutex_lock(&fs_list_lock);
369 fs_handle_t fs_handle;
370recheck:
371 fs_handle = fs_name_to_handle(instance, fs_name, false);
372 if (!fs_handle) {
373 if (flags & IPC_FLAG_BLOCKING) {
374 fibril_condvar_wait(&fs_list_cv, &fs_list_lock);
375 goto recheck;
376 }
377
378 fibril_mutex_unlock(&fs_list_lock);
379 async_answer_0(callid, ENOENT);
380 async_answer_0(rid, ENOENT);
381 free(mp);
382 free(fs_name);
383 free(opts);
384 return;
385 }
386 fibril_mutex_unlock(&fs_list_lock);
387
388 /* Add the filesystem info to the list of mounted filesystems */
389 mtab_ent_t *mtab_ent = malloc(sizeof(mtab_ent_t));
390 if (!mtab_ent) {
391 async_answer_0(callid, ENOMEM);
392 async_answer_0(rid, ENOMEM);
393 free(mp);
394 free(fs_name);
395 free(opts);
396 return;
397 }
398
399 /* Do the mount */
400 rc = vfs_mount_internal(rid, service_id, fs_handle, mp, opts);
401 if (rc != EOK) {
402 async_answer_0(callid, ENOTSUP);
403 async_answer_0(rid, ENOTSUP);
404 free(mtab_ent);
405 free(mp);
406 free(opts);
407 free(fs_name);
408 return;
409 }
410
411 /* Add the filesystem info to the list of mounted filesystems */
412
413 str_cpy(mtab_ent->mp, MAX_PATH_LEN, mp);
414 str_cpy(mtab_ent->fs_name, FS_NAME_MAXLEN, fs_name);
415 str_cpy(mtab_ent->opts, MAX_MNTOPTS_LEN, opts);
416 mtab_ent->instance = instance;
417 mtab_ent->service_id = service_id;
418
419 link_initialize(&mtab_ent->link);
420
421 fibril_mutex_lock(&mtab_list_lock);
422 list_append(&mtab_ent->link, &mtab_list);
423 mtab_size++;
424 fibril_mutex_unlock(&mtab_list_lock);
425
426 free(mp);
427 free(fs_name);
428 free(opts);
429
430 /* Acknowledge that we know fs_name. */
431 async_answer_0(callid, EOK);
432}
433
434void vfs_unmount(ipc_callid_t rid, ipc_call_t *request)
435{
436 int rc;
437 char *mp;
438 vfs_lookup_res_t mp_res;
439 vfs_lookup_res_t mr_res;
440 vfs_node_t *mr_node;
441 async_exch_t *exch;
442
443 /*
444 * Receive the mount point path.
445 */
446 rc = async_data_write_accept((void **) &mp, true, 0, MAX_PATH_LEN,
447 0, NULL);
448 if (rc != EOK)
449 async_answer_0(rid, rc);
450
451 /*
452 * Taking the namespace lock will do two things for us. First, it will
453 * prevent races with other lookup operations. Second, it will stop new
454 * references to already existing VFS nodes and creation of new VFS
455 * nodes. This is because new references are added as a result of some
456 * lookup operation or at least of some operation which is protected by
457 * the namespace lock.
458 */
459 fibril_rwlock_write_lock(&namespace_rwlock);
460
461 /*
462 * Lookup the mounted root and instantiate it.
463 */
464 rc = vfs_lookup_internal(root, mp, 0, &mr_res);
465 if (rc != EOK) {
466 fibril_rwlock_write_unlock(&namespace_rwlock);
467 free(mp);
468 async_answer_0(rid, rc);
469 return;
470 }
471 mr_node = vfs_node_get(&mr_res);
472 if (!mr_node) {
473 fibril_rwlock_write_unlock(&namespace_rwlock);
474 free(mp);
475 async_answer_0(rid, ENOMEM);
476 return;
477 }
478
479 /*
480 * Count the total number of references for the mounted file system. We
481 * are expecting at least two. One which we got above and one which we
482 * got when the file system was mounted. If we find more, it means that
483 * the file system cannot be gracefully unmounted at the moment because
484 * someone is working with it.
485 */
486 if (vfs_nodes_refcount_sum_get(mr_node->fs_handle,
487 mr_node->service_id) != 2) {
488 fibril_rwlock_write_unlock(&namespace_rwlock);
489 vfs_node_put(mr_node);
490 free(mp);
491 async_answer_0(rid, EBUSY);
492 return;
493 }
494
495 if (str_cmp(mp, "/") == 0) {
496
497 /*
498 * Unmounting the root file system.
499 *
500 * In this case, there is no mount point node and we send
501 * VFS_OUT_UNMOUNTED directly to the mounted file system.
502 */
503
504 exch = vfs_exchange_grab(mr_node->fs_handle);
505 rc = async_req_1_0(exch, VFS_OUT_UNMOUNTED,
506 mr_node->service_id);
507 vfs_exchange_release(exch);
508
509 if (rc != EOK) {
510 fibril_rwlock_write_unlock(&namespace_rwlock);
511 free(mp);
512 vfs_node_put(mr_node);
513 async_answer_0(rid, rc);
514 return;
515 }
516
517 root = NULL;
518 } else {
519
520 /*
521 * Unmounting a non-root file system.
522 *
523 * We have a regular mount point node representing the parent
524 * file system, so we delegate the operation to it.
525 */
526
527 rc = vfs_lookup_internal(root, mp, L_MP, &mp_res);
528 if (rc != EOK) {
529 fibril_rwlock_write_unlock(&namespace_rwlock);
530 free(mp);
531 vfs_node_put(mr_node);
532 async_answer_0(rid, rc);
533 return;
534 }
535
536 vfs_node_t *mp_node = vfs_node_get(&mp_res);
537 if (!mp_node) {
538 fibril_rwlock_write_unlock(&namespace_rwlock);
539 free(mp);
540 vfs_node_put(mr_node);
541 async_answer_0(rid, ENOMEM);
542 return;
543 }
544
545 exch = vfs_exchange_grab(mp_node->fs_handle);
546 rc = async_req_2_0(exch, VFS_OUT_UNMOUNT,
547 mp_node->service_id, mp_node->index);
548 vfs_exchange_release(exch);
549
550 if (rc != EOK) {
551 fibril_rwlock_write_unlock(&namespace_rwlock);
552 free(mp);
553 vfs_node_put(mp_node);
554 vfs_node_put(mr_node);
555 async_answer_0(rid, rc);
556 return;
557 }
558
559 /* Drop the reference we got above. */
560 vfs_node_put(mp_node);
561 /* Drop the reference from when the file system was mounted. */
562 vfs_node_put(mp_node);
563 }
564
565 /*
566 * All went well, the mounted file system was successfully unmounted.
567 * The only thing left is to forget the unmounted root VFS node.
568 */
569 vfs_node_forget(mr_node);
570 fibril_rwlock_write_unlock(&namespace_rwlock);
571
572 fibril_mutex_lock(&mtab_list_lock);
573
574 int found = 0;
575
576 list_foreach(mtab_list, cur) {
577 mtab_ent_t *mtab_ent = list_get_instance(cur, mtab_ent_t,
578 link);
579
580 if (str_cmp(mtab_ent->mp, mp) == 0) {
581 list_remove(&mtab_ent->link);
582 mtab_size--;
583 free(mtab_ent);
584 found = 1;
585 break;
586 }
587 }
588 assert(found);
589 fibril_mutex_unlock(&mtab_list_lock);
590
591 free(mp);
592
593 async_answer_0(rid, EOK);
594}
595
596static inline bool walk_flags_valid(int flags)
597{
598 if ((flags&~WALK_ALL_FLAGS) != 0) {
599 return false;
600 }
601 if ((flags&WALK_MAY_CREATE) && (flags&WALK_MUST_CREATE)) {
602 return false;
603 }
604 if ((flags&WALK_REGULAR) && (flags&WALK_DIRECTORY)) {
605 return false;
606 }
607 if ((flags&WALK_MAY_CREATE) || (flags&WALK_MUST_CREATE)) {
608 if (!(flags&WALK_DIRECTORY) && !(flags&WALK_REGULAR)) {
609 return false;
610 }
611 }
612 return true;
613}
614
615static inline int walk_lookup_flags(int flags)
616{
617 int lflags = 0;
618 if (flags&WALK_MAY_CREATE || flags&WALK_MUST_CREATE) {
619 lflags |= L_CREATE;
620 }
621 if (flags&WALK_MUST_CREATE) {
622 lflags |= L_EXCLUSIVE;
623 }
624 if (flags&WALK_REGULAR) {
625 lflags |= L_FILE;
626 }
627 if (flags&WALK_DIRECTORY) {
628 lflags |= L_DIRECTORY;
629 }
630 return lflags;
631}
632
633void vfs_walk(ipc_callid_t rid, ipc_call_t *request)
634{
635 /*
636 * Parent is our relative root for file lookup.
637 * For defined flags, see <ipc/vfs.h>.
638 */
639 int parentfd = IPC_GET_ARG1(*request);
640 int flags = IPC_GET_ARG2(*request);
641
642 if (!walk_flags_valid(flags)) {
643 async_answer_0(rid, EINVAL);
644 return;
645 }
646
647 char *path;
648 int rc = async_data_write_accept((void **)&path, true, 0, 0, 0, NULL);
649
650 /* Lookup the file structure corresponding to the file descriptor. */
651 vfs_file_t *parent = NULL;
652 vfs_node_t *parent_node = root;
653 // TODO: Client-side root.
654 if (parentfd != -1) {
655 parent = vfs_file_get(parentfd);
656 if (!parent) {
657 free(path);
658 async_answer_0(rid, EBADF);
659 return;
660 }
661 parent_node = parent->node;
662 }
663
664 fibril_rwlock_read_lock(&namespace_rwlock);
665
666 vfs_lookup_res_t lr;
667 rc = vfs_lookup_internal(parent_node, path, walk_lookup_flags(flags), &lr);
668 free(path);
669
670 if (rc != EOK) {
671 fibril_rwlock_read_unlock(&namespace_rwlock);
672 if (parent) {
673 vfs_file_put(parent);
674 }
675 async_answer_0(rid, rc);
676 return;
677 }
678
679 vfs_node_t *node = vfs_node_get(&lr);
680
681 int fd = vfs_fd_alloc(false);
682 if (fd < 0) {
683 vfs_node_put(node);
684 if (parent) {
685 vfs_file_put(parent);
686 }
687 async_answer_0(rid, fd);
688 return;
689 }
690
691 vfs_file_t *file = vfs_file_get(fd);
692 assert(file != NULL);
693
694 file->node = node;
695 if (parent) {
696 file->permissions = parent->permissions;
697 } else {
698 file->permissions = MODE_READ | MODE_WRITE | MODE_APPEND;
699 }
700 file->open_read = false;
701 file->open_write = false;
702
703 vfs_file_put(file);
704 if (parent) {
705 vfs_file_put(parent);
706 }
707
708 fibril_rwlock_read_unlock(&namespace_rwlock);
709
710 async_answer_1(rid, EOK, fd);
711}
712
713void vfs_open2(ipc_callid_t rid, ipc_call_t *request)
714{
715 int fd = IPC_GET_ARG1(*request);
716 int flags = IPC_GET_ARG2(*request);
717
718 if (flags == 0) {
719 async_answer_0(rid, EINVAL);
720 return;
721 }
722
723 vfs_file_t *file = vfs_file_get(fd);
724 if (!file) {
725 async_answer_0(rid, EBADF);
726 return;
727 }
728
729 if ((flags & ~file->permissions) != 0) {
730 vfs_file_put(file);
731 async_answer_0(rid, EPERM);
732 return;
733 }
734
735 file->open_read = (flags & MODE_READ) != 0;
736 file->open_write = (flags & (MODE_WRITE | MODE_APPEND)) != 0;
737 file->append = (flags & MODE_APPEND) != 0;
738
739 if (!file->open_read && !file->open_write) {
740 vfs_file_put(file);
741 async_answer_0(rid, EINVAL);
742 return;
743 }
744
745 if (file->node->type == VFS_NODE_DIRECTORY && file->open_write) {
746 file->open_read = file->open_write = false;
747 vfs_file_put(file);
748 async_answer_0(rid, EINVAL);
749 return;
750 }
751
752 int rc = vfs_open_node_remote(file->node);
753 if (rc != EOK) {
754 file->open_read = file->open_write = false;
755 vfs_file_put(file);
756 async_answer_0(rid, rc);
757 return;
758 }
759
760 vfs_file_put(file);
761 async_answer_0(rid, EOK);
762}
763
764void vfs_sync(ipc_callid_t rid, ipc_call_t *request)
765{
766 int fd = IPC_GET_ARG1(*request);
767
768 /* Lookup the file structure corresponding to the file descriptor. */
769 vfs_file_t *file = vfs_file_get(fd);
770 if (!file) {
771 async_answer_0(rid, ENOENT);
772 return;
773 }
774
775 /*
776 * Lock the open file structure so that no other thread can manipulate
777 * the same open file at a time.
778 */
779 fibril_mutex_lock(&file->lock);
780 async_exch_t *fs_exch = vfs_exchange_grab(file->node->fs_handle);
781
782 /* Make a VFS_OUT_SYMC request at the destination FS server. */
783 aid_t msg;
784 ipc_call_t answer;
785 msg = async_send_2(fs_exch, VFS_OUT_SYNC, file->node->service_id,
786 file->node->index, &answer);
787
788 vfs_exchange_release(fs_exch);
789
790 /* Wait for reply from the FS server. */
791 sysarg_t rc;
792 async_wait_for(msg, &rc);
793
794 fibril_mutex_unlock(&file->lock);
795
796 vfs_file_put(file);
797 async_answer_0(rid, rc);
798}
799
800void vfs_close(ipc_callid_t rid, ipc_call_t *request)
801{
802 int fd = IPC_GET_ARG1(*request);
803 int ret = vfs_fd_free(fd);
804 async_answer_0(rid, ret);
805}
806
807static void vfs_rdwr(ipc_callid_t rid, ipc_call_t *request, bool read)
808{
809 /*
810 * The following code strongly depends on the fact that the files data
811 * structure can be only accessed by a single fibril and all file
812 * operations are serialized (i.e. the reads and writes cannot
813 * interleave and a file cannot be closed while it is being read).
814 *
815 * Additional synchronization needs to be added once the table of
816 * open files supports parallel access!
817 */
818
819 int fd = IPC_GET_ARG1(*request);
820
821 /* Lookup the file structure corresponding to the file descriptor. */
822 vfs_file_t *file = vfs_file_get(fd);
823 if (!file) {
824 async_answer_0(rid, ENOENT);
825 return;
826 }
827
828 /*
829 * Lock the open file structure so that no other thread can manipulate
830 * the same open file at a time.
831 */
832 fibril_mutex_lock(&file->lock);
833
834 if ((read && !file->open_read) || (!read && !file->open_write)) {
835 fibril_mutex_unlock(&file->lock);
836 async_answer_0(rid, EINVAL);
837 return;
838 }
839
840 vfs_info_t *fs_info = fs_handle_to_info(file->node->fs_handle);
841 assert(fs_info);
842
843 /*
844 * Lock the file's node so that no other client can read/write to it at
845 * the same time unless the FS supports concurrent reads/writes and its
846 * write implementation does not modify the file size.
847 */
848 if ((read) ||
849 ((fs_info->concurrent_read_write) && (fs_info->write_retains_size)))
850 fibril_rwlock_read_lock(&file->node->contents_rwlock);
851 else
852 fibril_rwlock_write_lock(&file->node->contents_rwlock);
853
854 if (file->node->type == VFS_NODE_DIRECTORY) {
855 /*
856 * Make sure that no one is modifying the namespace
857 * while we are in readdir().
858 */
859 assert(read);
860 fibril_rwlock_read_lock(&namespace_rwlock);
861 }
862
863 async_exch_t *fs_exch = vfs_exchange_grab(file->node->fs_handle);
864
865 /*
866 * Make a VFS_READ/VFS_WRITE request at the destination FS server
867 * and forward the IPC_M_DATA_READ/IPC_M_DATA_WRITE request to the
868 * destination FS server. The call will be routed as if sent by
869 * ourselves. Note that call arguments are immutable in this case so we
870 * don't have to bother.
871 */
872 sysarg_t rc;
873 ipc_call_t answer;
874 if (read) {
875 rc = async_data_read_forward_4_1(fs_exch, VFS_OUT_READ,
876 file->node->service_id, file->node->index,
877 LOWER32(file->pos), UPPER32(file->pos), &answer);
878 } else {
879 if (file->append)
880 file->pos = vfs_node_get_size(file->node);
881
882 rc = async_data_write_forward_4_1(fs_exch, VFS_OUT_WRITE,
883 file->node->service_id, file->node->index,
884 LOWER32(file->pos), UPPER32(file->pos), &answer);
885 }
886
887 vfs_exchange_release(fs_exch);
888
889 size_t bytes = IPC_GET_ARG1(answer);
890
891 if (file->node->type == VFS_NODE_DIRECTORY) {
892 fibril_rwlock_read_unlock(&namespace_rwlock);
893 }
894
895 /* Unlock the VFS node. */
896 if ((read) ||
897 ((fs_info->concurrent_read_write) && (fs_info->write_retains_size)))
898 fibril_rwlock_read_unlock(&file->node->contents_rwlock);
899 else {
900 /* Update the cached version of node's size. */
901 if (rc == EOK)
902 file->node->size = MERGE_LOUP32(IPC_GET_ARG2(answer),
903 IPC_GET_ARG3(answer));
904 fibril_rwlock_write_unlock(&file->node->contents_rwlock);
905 }
906
907 /* Update the position pointer and unlock the open file. */
908 if (rc == EOK)
909 file->pos += bytes;
910 fibril_mutex_unlock(&file->lock);
911 vfs_file_put(file);
912
913 /*
914 * FS server's reply is the final result of the whole operation we
915 * return to the client.
916 */
917 async_answer_1(rid, rc, bytes);
918}
919
920void vfs_read(ipc_callid_t rid, ipc_call_t *request)
921{
922 vfs_rdwr(rid, request, true);
923}
924
925void vfs_write(ipc_callid_t rid, ipc_call_t *request)
926{
927 vfs_rdwr(rid, request, false);
928}
929
930void vfs_seek(ipc_callid_t rid, ipc_call_t *request)
931{
932 int fd = (int) IPC_GET_ARG1(*request);
933 off64_t off = (off64_t) MERGE_LOUP32(IPC_GET_ARG2(*request),
934 IPC_GET_ARG3(*request));
935 int whence = (int) IPC_GET_ARG4(*request);
936
937 /* Lookup the file structure corresponding to the file descriptor. */
938 vfs_file_t *file = vfs_file_get(fd);
939 if (!file) {
940 async_answer_0(rid, ENOENT);
941 return;
942 }
943
944 fibril_mutex_lock(&file->lock);
945
946 off64_t newoff;
947 switch (whence) {
948 case SEEK_SET:
949 if (off >= 0) {
950 file->pos = (aoff64_t) off;
951 fibril_mutex_unlock(&file->lock);
952 vfs_file_put(file);
953 async_answer_1(rid, EOK, off);
954 return;
955 }
956 break;
957 case SEEK_CUR:
958 if ((off >= 0) && (file->pos + off < file->pos)) {
959 fibril_mutex_unlock(&file->lock);
960 vfs_file_put(file);
961 async_answer_0(rid, EOVERFLOW);
962 return;
963 }
964
965 if ((off < 0) && (file->pos < (aoff64_t) -off)) {
966 fibril_mutex_unlock(&file->lock);
967 vfs_file_put(file);
968 async_answer_0(rid, EOVERFLOW);
969 return;
970 }
971
972 file->pos += off;
973 newoff = (file->pos > OFF64_MAX) ? OFF64_MAX : file->pos;
974
975 fibril_mutex_unlock(&file->lock);
976 vfs_file_put(file);
977 async_answer_2(rid, EOK, LOWER32(newoff),
978 UPPER32(newoff));
979 return;
980 case SEEK_END:
981 fibril_rwlock_read_lock(&file->node->contents_rwlock);
982 aoff64_t size = vfs_node_get_size(file->node);
983
984 if ((off >= 0) && (size + off < size)) {
985 fibril_rwlock_read_unlock(&file->node->contents_rwlock);
986 fibril_mutex_unlock(&file->lock);
987 vfs_file_put(file);
988 async_answer_0(rid, EOVERFLOW);
989 return;
990 }
991
992 if ((off < 0) && (size < (aoff64_t) -off)) {
993 fibril_rwlock_read_unlock(&file->node->contents_rwlock);
994 fibril_mutex_unlock(&file->lock);
995 vfs_file_put(file);
996 async_answer_0(rid, EOVERFLOW);
997 return;
998 }
999
1000 file->pos = size + off;
1001 newoff = (file->pos > OFF64_MAX) ? OFF64_MAX : file->pos;
1002
1003 fibril_rwlock_read_unlock(&file->node->contents_rwlock);
1004 fibril_mutex_unlock(&file->lock);
1005 vfs_file_put(file);
1006 async_answer_2(rid, EOK, LOWER32(newoff), UPPER32(newoff));
1007 return;
1008 }
1009
1010 fibril_mutex_unlock(&file->lock);
1011 vfs_file_put(file);
1012 async_answer_0(rid, EINVAL);
1013}
1014
1015int vfs_truncate_internal(fs_handle_t fs_handle, service_id_t service_id,
1016 fs_index_t index, aoff64_t size)
1017{
1018 async_exch_t *exch = vfs_exchange_grab(fs_handle);
1019 sysarg_t rc = async_req_4_0(exch, VFS_OUT_TRUNCATE,
1020 (sysarg_t) service_id, (sysarg_t) index, LOWER32(size),
1021 UPPER32(size));
1022 vfs_exchange_release(exch);
1023
1024 return (int) rc;
1025}
1026
1027void vfs_truncate(ipc_callid_t rid, ipc_call_t *request)
1028{
1029 int fd = IPC_GET_ARG1(*request);
1030 aoff64_t size = (aoff64_t) MERGE_LOUP32(IPC_GET_ARG2(*request),
1031 IPC_GET_ARG3(*request));
1032 int rc;
1033
1034 vfs_file_t *file = vfs_file_get(fd);
1035 if (!file) {
1036 async_answer_0(rid, ENOENT);
1037 return;
1038 }
1039 fibril_mutex_lock(&file->lock);
1040
1041 fibril_rwlock_write_lock(&file->node->contents_rwlock);
1042 rc = vfs_truncate_internal(file->node->fs_handle,
1043 file->node->service_id, file->node->index, size);
1044 if (rc == EOK)
1045 file->node->size = size;
1046 fibril_rwlock_write_unlock(&file->node->contents_rwlock);
1047
1048 fibril_mutex_unlock(&file->lock);
1049 vfs_file_put(file);
1050 async_answer_0(rid, (sysarg_t)rc);
1051}
1052
1053void vfs_fstat(ipc_callid_t rid, ipc_call_t *request)
1054{
1055 int fd = IPC_GET_ARG1(*request);
1056 sysarg_t rc;
1057
1058 vfs_file_t *file = vfs_file_get(fd);
1059 if (!file) {
1060 async_answer_0(rid, ENOENT);
1061 return;
1062 }
1063
1064 ipc_callid_t callid;
1065 if (!async_data_read_receive(&callid, NULL)) {
1066 vfs_file_put(file);
1067 async_answer_0(callid, EINVAL);
1068 async_answer_0(rid, EINVAL);
1069 return;
1070 }
1071
1072 fibril_mutex_lock(&file->lock);
1073
1074 async_exch_t *exch = vfs_exchange_grab(file->node->fs_handle);
1075
1076 aid_t msg;
1077 msg = async_send_3(exch, VFS_OUT_STAT, file->node->service_id,
1078 file->node->index, true, NULL);
1079 async_forward_fast(callid, exch, 0, 0, 0, IPC_FF_ROUTE_FROM_ME);
1080
1081 vfs_exchange_release(exch);
1082
1083 async_wait_for(msg, &rc);
1084
1085 fibril_mutex_unlock(&file->lock);
1086 vfs_file_put(file);
1087 async_answer_0(rid, rc);
1088}
1089
1090static void out_destroy(vfs_triplet_t *file)
1091{
1092 async_exch_t *exch = vfs_exchange_grab(file->fs_handle);
1093 async_msg_2(exch, VFS_OUT_DESTROY,
1094 (sysarg_t) file->service_id, (sysarg_t) file->index);
1095 vfs_exchange_release(exch);
1096}
1097
1098void vfs_unlink2(ipc_callid_t rid, ipc_call_t *request)
1099{
1100 int rc;
1101 char *path;
1102 vfs_file_t *parent = NULL;
1103 vfs_file_t *expect = NULL;
1104 vfs_node_t *parent_node = root;
1105
1106 int parentfd = IPC_GET_ARG1(*request);
1107 int expectfd = IPC_GET_ARG2(*request);
1108 int wflag = IPC_GET_ARG3(*request);
1109
1110 rc = async_data_write_accept((void **) &path, true, 0, 0, 0, NULL);
1111 if (rc != EOK) {
1112 async_answer_0(rid, rc);
1113 return;
1114 }
1115
1116 fibril_rwlock_write_lock(&namespace_rwlock);
1117
1118 int lflag = (wflag&WALK_DIRECTORY) ? L_DIRECTORY: 0;
1119
1120 if (parentfd >= 0) {
1121 parent = vfs_file_get(parentfd);
1122 if (!parent) {
1123 rc = ENOENT;
1124 goto exit;
1125 }
1126 parent_node = parent->node;
1127 }
1128
1129 if (expectfd >= 0) {
1130 expect = vfs_file_get(expectfd);
1131 if (!expect) {
1132 rc = ENOENT;
1133 goto exit;
1134 }
1135
1136 vfs_lookup_res_t lr;
1137 rc = vfs_lookup_internal(parent_node, path, lflag, &lr);
1138 if (rc != EOK) {
1139 goto exit;
1140 }
1141
1142 if (__builtin_memcmp(&lr.triplet, expect->node, sizeof(vfs_triplet_t)) != 0) {
1143 rc = ENOENT;
1144 goto exit;
1145 }
1146
1147 vfs_file_put(expect);
1148 expect = NULL;
1149 }
1150
1151 vfs_lookup_res_t lr;
1152 rc = vfs_lookup_internal(parent_node, path, lflag | L_UNLINK, &lr);
1153 if (rc != EOK) {
1154 goto exit;
1155 }
1156
1157 /* If the node is not held by anyone, try to destroy it. */
1158 if (vfs_node_peek(&lr) == NULL) {
1159 out_destroy(&lr.triplet);
1160 }
1161
1162exit:
1163 if (path) {
1164 free(path);
1165 }
1166 if (parent) {
1167 vfs_file_put(parent);
1168 }
1169 if (expect) {
1170 vfs_file_put(expect);
1171 }
1172 fibril_rwlock_write_unlock(&namespace_rwlock);
1173 async_answer_0(rid, rc);
1174}
1175
1176static size_t shared_path(char *a, char *b)
1177{
1178 size_t res = 0;
1179
1180 while (a[res] == b[res] && a[res] != 0) {
1181 res++;
1182 }
1183
1184 if (a[res] == b[res]) {
1185 return res;
1186 }
1187
1188 res--;
1189 while (a[res] != '/') {
1190 res--;
1191 }
1192 return res;
1193}
1194
1195static int vfs_rename_internal(vfs_node_t *base, char *old, char *new)
1196{
1197 assert(base != NULL);
1198 assert(old != NULL);
1199 assert(new != NULL);
1200
1201 vfs_lookup_res_t base_lr;
1202 vfs_lookup_res_t old_lr;
1203 vfs_lookup_res_t new_lr_orig;
1204 bool orig_unlinked = false;
1205
1206 int rc;
1207
1208 size_t shared = shared_path(old, new);
1209
1210 /* Do not allow one path to be a prefix of the other. */
1211 if (old[shared] == 0 || new[shared] == 0) {
1212 return EINVAL;
1213 }
1214 assert(old[shared] == '/');
1215 assert(new[shared] == '/');
1216
1217 fibril_rwlock_write_lock(&namespace_rwlock);
1218
1219 /* Resolve the shared portion of the path first. */
1220 if (shared != 0) {
1221 old[shared] = 0;
1222 rc = vfs_lookup_internal(base, old, L_DIRECTORY, &base_lr);
1223 if (rc != EOK) {
1224 fibril_rwlock_write_unlock(&namespace_rwlock);
1225 return rc;
1226 }
1227
1228 base = vfs_node_get(&base_lr);
1229 old[shared] = '/';
1230 old += shared;
1231 new += shared;
1232 } else {
1233 vfs_node_addref(base);
1234 }
1235
1236
1237 rc = vfs_lookup_internal(base, new, L_UNLINK | L_DISABLE_MOUNTS, &new_lr_orig);
1238 if (rc == EOK) {
1239 orig_unlinked = true;
1240 } else if (rc != ENOENT) {
1241 vfs_node_put(base);
1242 fibril_rwlock_write_unlock(&namespace_rwlock);
1243 return rc;
1244 }
1245
1246 rc = vfs_lookup_internal(base, old, L_UNLINK | L_DISABLE_MOUNTS, &old_lr);
1247 if (rc != EOK) {
1248 if (orig_unlinked) {
1249 vfs_link_internal(base, new, &new_lr_orig.triplet);
1250 }
1251 vfs_node_put(base);
1252 fibril_rwlock_write_unlock(&namespace_rwlock);
1253 return rc;
1254 }
1255
1256 rc = vfs_link_internal(base, new, &old_lr.triplet);
1257 if (rc != EOK) {
1258 vfs_link_internal(base, old, &old_lr.triplet);
1259 if (orig_unlinked) {
1260 vfs_link_internal(base, new, &new_lr_orig.triplet);
1261 }
1262 vfs_node_put(base);
1263 fibril_rwlock_write_unlock(&namespace_rwlock);
1264 return rc;
1265 }
1266
1267 /* If the node is not held by anyone, try to destroy it. */
1268 if (orig_unlinked && vfs_node_peek(&new_lr_orig) == NULL) {
1269 out_destroy(&new_lr_orig.triplet);
1270 }
1271
1272 vfs_node_put(base);
1273 fibril_rwlock_write_unlock(&namespace_rwlock);
1274 return EOK;
1275}
1276
1277void vfs_rename(ipc_callid_t rid, ipc_call_t *request)
1278{
1279 /* The common base directory. */
1280 int basefd;
1281 char *old = NULL;
1282 char *new = NULL;
1283 vfs_file_t *base = NULL;
1284 int rc;
1285
1286 basefd = IPC_GET_ARG1(*request);
1287
1288 /* Retrieve the old path. */
1289 rc = async_data_write_accept((void **) &old, true, 0, 0, 0, NULL);
1290 if (rc != EOK) {
1291 goto out;
1292 }
1293
1294 /* Retrieve the new path. */
1295 rc = async_data_write_accept((void **) &new, true, 0, 0, 0, NULL);
1296 if (rc != EOK) {
1297 goto out;
1298 }
1299
1300 size_t olen;
1301 size_t nlen;
1302 char *oldc = canonify(old, &olen);
1303 char *newc = canonify(new, &nlen);
1304
1305 if ((!oldc) || (!newc)) {
1306 rc = EINVAL;
1307 goto out;
1308 }
1309
1310 assert(oldc[olen] == '\0');
1311 assert(newc[nlen] == '\0');
1312
1313 /* Lookup the file structure corresponding to the file descriptor. */
1314 vfs_node_t *base_node = root;
1315 // TODO: Client-side root.
1316 if (basefd != -1) {
1317 base = vfs_file_get(basefd);
1318 if (!base) {
1319 rc = EBADF;
1320 goto out;
1321 }
1322 base_node = base->node;
1323 }
1324
1325 rc = vfs_rename_internal(base_node, oldc, newc);
1326
1327out:
1328 async_answer_0(rid, rc);
1329
1330 if (old) {
1331 free(old);
1332 }
1333 if (new) {
1334 free(new);
1335 }
1336 if (base) {
1337 vfs_file_put(base);
1338 }
1339}
1340
1341void vfs_dup(ipc_callid_t rid, ipc_call_t *request)
1342{
1343 int oldfd = IPC_GET_ARG1(*request);
1344 int newfd = IPC_GET_ARG2(*request);
1345
1346 /* If the file descriptors are the same, do nothing. */
1347 if (oldfd == newfd) {
1348 async_answer_1(rid, EOK, newfd);
1349 return;
1350 }
1351
1352 /* Lookup the file structure corresponding to oldfd. */
1353 vfs_file_t *oldfile = vfs_file_get(oldfd);
1354 if (!oldfile) {
1355 async_answer_0(rid, EBADF);
1356 return;
1357 }
1358
1359 /*
1360 * Lock the open file structure so that no other thread can manipulate
1361 * the same open file at a time.
1362 */
1363 fibril_mutex_lock(&oldfile->lock);
1364
1365 /* Make sure newfd is closed. */
1366 (void) vfs_fd_free(newfd);
1367
1368 /* Assign the old file to newfd. */
1369 int ret = vfs_fd_assign(oldfile, newfd);
1370 fibril_mutex_unlock(&oldfile->lock);
1371 vfs_file_put(oldfile);
1372
1373 if (ret != EOK)
1374 async_answer_0(rid, ret);
1375 else
1376 async_answer_1(rid, EOK, newfd);
1377}
1378
1379void vfs_wait_handle(ipc_callid_t rid, ipc_call_t *request)
1380{
1381 int fd = vfs_wait_handle_internal();
1382 async_answer_1(rid, EOK, fd);
1383}
1384
1385void vfs_get_mtab(ipc_callid_t rid, ipc_call_t *request)
1386{
1387 ipc_callid_t callid;
1388 ipc_call_t data;
1389 sysarg_t rc = EOK;
1390 size_t len;
1391
1392 fibril_mutex_lock(&mtab_list_lock);
1393
1394 /* Send to the caller the number of mounted filesystems */
1395 callid = async_get_call(&data);
1396 if (IPC_GET_IMETHOD(data) != VFS_IN_PING) {
1397 rc = ENOTSUP;
1398 async_answer_0(callid, rc);
1399 goto exit;
1400 }
1401 async_answer_1(callid, EOK, mtab_size);
1402
1403 list_foreach(mtab_list, cur) {
1404 mtab_ent_t *mtab_ent = list_get_instance(cur, mtab_ent_t,
1405 link);
1406
1407 rc = ENOTSUP;
1408
1409 if (!async_data_read_receive(&callid, &len)) {
1410 async_answer_0(callid, rc);
1411 goto exit;
1412 }
1413
1414 (void) async_data_read_finalize(callid, mtab_ent->mp,
1415 str_size(mtab_ent->mp));
1416
1417 if (!async_data_read_receive(&callid, &len)) {
1418 async_answer_0(callid, rc);
1419 goto exit;
1420 }
1421
1422 (void) async_data_read_finalize(callid, mtab_ent->opts,
1423 str_size(mtab_ent->opts));
1424
1425 if (!async_data_read_receive(&callid, &len)) {
1426 async_answer_0(callid, rc);
1427 goto exit;
1428 }
1429
1430 (void) async_data_read_finalize(callid, mtab_ent->fs_name,
1431 str_size(mtab_ent->fs_name));
1432
1433 callid = async_get_call(&data);
1434
1435 if (IPC_GET_IMETHOD(data) != VFS_IN_PING) {
1436 async_answer_0(callid, rc);
1437 goto exit;
1438 }
1439
1440 rc = EOK;
1441 async_answer_2(callid, rc, mtab_ent->instance,
1442 mtab_ent->service_id);
1443 }
1444
1445exit:
1446 fibril_mutex_unlock(&mtab_list_lock);
1447 async_answer_0(rid, rc);
1448}
1449
1450/**
1451 * @}
1452 */
Note: See TracBrowser for help on using the repository browser.